Associations between Depression, Anxiety, Fatigue, and Learning Motivating Factors in e-Learning-Based Computer Programming Education

Quarantines imposed due to COVID-19 have forced the rapid implementation of e-learning, but also increased the rates of anxiety, depression, and fatigue, which relate to dramatically diminished e-learning motivation. Thus, it was deemed significant to identify e-learning motivating factors related to mental health. Furthermore, because computer programming skills are among the core competencies that professionals are expected to possess in the era of rapid technology development, it was also considered important to identify the factors relating to computer programming learning. Thus, this study applied the Learning Motivating Factors Questionnaire, the Patient Health Questionnaire-9 (PHQ-9), the Generalized Anxiety Disorder Scale-7 (GAD-7), and the Multidimensional Fatigue Inventory-20 (MFI-20) instruments. The sample consisted of 444 e-learners, including 189 computer programming e-learners. The results revealed that higher scores of individual attitude and expectation, challenging goals, clear direction, social pressure, and competition significantly varied across depression categories. The scores of challenging goals, and social pressure and competition, significantly varied across anxiety categories. The scores of individual attitude and expectation, challenging goals, and social pressure and competition significantly varied across general fatigue categories. In the group of computer programming e-learners: challenging goals predicted decreased anxiety; clear direction and challenging goals predicted decreased depression; individual attitude and expectation predicted diminished general fatigue; and challenging goals and punishment predicted diminished mental fatigue. Challenging goals statistically significantly predicted lower mental fatigue, and mental fatigue statistically significantly predicted depression and anxiety in both sample groups.


Introduction
Computer programming skills and computational literacy are among the core competencies that professionals are expected to possess in the era of rapid technology development and a fast-changing work environment [1].
Research indicates that learners of various ages are motivated to acquire computer programming skills because computer programming generates cognitive benefits [2] and may ensure more extensive possibilities of work or entrepreneurship [3]. For example, Python is a general-purpose programming language that empowers developers to use several different programming styles when creating programs. According to HackerRank, Python is the second-most in-demand programming language, and over 50% of hiring managers seek candidates who know this language [4]. The programming languages of Python, Java, HTML, CSS, SQL, NoSQL, C#, Rust, Perl, and Go are the most demanded in 2021, and courses aimed at learning these programming languages are in high demand [5].
Several studies have demonstrated the relationship between neuroticism, which reflects trait anxiety and depression [57], and learning achievements [60], evidencing that anxiety and depression lead to diminished performance. Research has also revealed that neuroticism is positively related to extrinsic academic motivation [61]. However, the importance of mental health in e-learning education is under-researched, despite attempts to conceptualize challenges in e-learning education [62][63][64].
Identifying challenges in e-learning education is especially significant because, as a result of the coronavirus outbreak, numerous countries face changes in many sectors, including education. Many students globally have been affected by the suspension of classes, and countries have faced the need for rapid implementation of e-learning. As a result, the question of learning motivation in e-learning is extremely important, particularly considering the observed statistically significant increase in the rates of anxiety and depression [65] and chronic fatigue [66].
Research suggests that university students were susceptible to developing depression and anxiety during the COVID-19 quarantine due to the psychologically challenging circumstances they faced each day [67,68]. Although anxiety levels may depend on coping strategies [69], and depression levels may depend on social support [70], social isolation during the COVID-19 pandemic can intensify feelings of worthlessness, and presumed dangers can increase worrying and anxiety, which combined may finally lead to depression and emotional exhaustion [71].
Several studies revealed that most students during the quarantine experienced increased anxiety and depression [70,72,73]. Furthermore, research found a significant relationship between students' satisfaction with e-learning and the prevalence of depression and anxiety symptoms [67]. Shockingly, during the COVID-19 pandemic, more than half of students met the diagnostic criteria of a generalized anxiety disorder (52%) and depression (63%) [74]. Moreover, although some mild anxiety was related to an increase in motivation [75], the states of severe anxiety and depression were related to dramatically diminished e-learning motivation, or even its absence [76].
To summarize, associations between learning motivating factors and depression, anxiety, and fatigue are under-researched, but there is some evidence that these states may relate to constrained learning motivation. From the perspective of cognitive processes, one of the significant factors contributing to the inhibition of learning motivation is subjective fatigue [77,78], which refers to a decline in mental efficiency and the accompanying feelings of weariness [79,80], and more generally represents a crucial element in the effort-regulation system as a whole [81]. By comparison, extrinsic motivation may lead to emotional exhaustion, and mental and general fatigue, but the paths and contributing factors are under-researched. Therefore, it is important to analyze the associations between the learning motivating factors and depression, anxiety, and fatigue in e-learning-based education. Furthermore, to identify the specific factors relating to computer programming e-learning, it is essential to analyze the differences between those who participate in e-learning-based computer learning courses and those who study in other university programs. Consequently, although this study targeted students enrolled in e-learning-based computer programming education, to reveal the specifics of programming e-learners, the comparative group consisted of university students who studied social sciences. Due to the COVID-19 pandemic, these students were studying remotely.
Thus, this study aimed to identify associations between depression, anxiety, fatigue, and learning motivating factors in e-learning-based education. Furthermore, this study aimed to compare the patterns in participants and non-participants of e-learning-based computer programming courses, and to identify the specific factors relating to computer programming e-learners.
Based on the previous research, we hypothesized that: Hypotheses 1 (H1). Lower scores of depression relate to higher scores of learning motivating factors; Hypotheses 2 (H2). Lower anxiety scores relate to higher scores of intrinsic, but not extrinsic, learning motivating factors; Hypotheses 3 (H3). Lower scores of general fatigue relate to higher scores of learning motivating factors; Hypotheses 4 (H4). Extrinsic learning motivating factors predict increased anxiety, and intrinsic learning motivating factors predict decreased anxiety; Hypotheses 5 (H5). Extrinsic learning motivating factors predict increased depression, and intrinsic learning motivating factors predict decreased depression; Hypotheses 6 (H6). Extrinsic learning motivating factors, anxiety, and depression predict increased general and mental fatigue, and intrinsic learning motivating factors predict decreased general and mental fatigue; Hypotheses 7 (H7). Associations exist between learning motivating factors, anxiety, depression, and fatigue, but they differ between participants and non-participants of e-learning-based computer programming courses.

Sample
In the full sample of 465 participants, a total of 444 participants had no missing data. Because the number of cases with missing values was small, we used listwise deletion of cases with missing values. Therefore, all analyses were conducted using a sample of 444 individuals. The sample's breakdown by gender was 32.7 percent male (n = 145) and 67.3 percent female (n = 299). The respondents' mean age was 25.19 years (standard deviation (SD) = 8.268, 95% confidence interval (CI) = 24.42, 25.97, age range = 18-57 years). A total of 189 (42.6%) of participants studied in e-learning-based computer programming courses organized by Turing College. Based on Ding, Velicer, and Harlow [82], a sample size within the range of 100 to 200 is sufficient for structural equation modeling [82]. Thus, the sample size of the computer programming e-learners in this study was satisfactory to obtain meaningful results in the applied statistical models.
The comparative group consisted of 255 (57.4%) respondents who studied social sciences at various Lithuanian Universities; however, these students were studying remotely due to the COVID-19 pandemic.
At the time of the research, both groups of e-learners (computer programming and social sciences) were undertaking their studies. E-learners were informed about the study by e-mail and provided their consent to participate in the research. Participation in the study was voluntary, and the participants did not receive any compensation. The procedure was administered online at https://www.psytest.online (accessed on 7 July 2021) and followed the General Data Protection Regulation (GDPR) guidelines. The study was approved by the Institutional Review Board of the Institute of Management and Psychology, based on the approval of the Biomedical Research Scientific-Ethics Committee at Klaipėda University (No. STIMC-BMTEK-P03, 12 April 2021).

Instruments
This study applied four instruments: the translated Lithuanian version of the Learning Motivating Factors Questionnaire [6,83], the translated Lithuanian version of the Patient Health Questionnaire-9 (PHQ-9) [84], the translated Lithuanian version of the Generalized Anxiety Disorder Scale (GAD-7) [85], and the translated Lithuanian version of the Multidimensional Fatigue Inventory (MFI-20) [86]. To ensure that the Lithuanian items corresponded as closely as possible to the English items, the original items of both instruments were translated into Lithuanian and back-translated.

The PHQ-9
To assess depression, we applied the Patient Health Questionnaire-9 (PHQ-9) [84], which is a 9 item self-reported measure used to assess depression severity and criteria for a major depressive episode (MDE). Items are assessed for symptoms of depression (e.g., "little interest or pleasure in doing things"), and response anchors range temporally from 0 (not at all) to 3 (nearly every day). Items are summed to create a severity score ranging from 0 to 27, with higher scores reflecting greater depression severity. Scores above 10 are considered to be in the depressive area [84]. The scores of the PHQ-9 are usually used as categorical variables due to their clinical relevance. In this study, we present only frequencies and ANOVA results based on the PHQ-9 categories. Due to the relatively small sample size, we used the PHQ-9 as a continuous variable to apply the SEM, because research has suggested that the PHQ-9 can also be examined as a continuous variable [84]. To apply the SEM with categorical variables, a considerably larger sample in each category is needed. Unfortunately, the number of participants in each category was insufficient for the chosen statistical models.

The GAD-7
To assess anxiety, we applied the Generalized Anxiety Disorder Scale (GAD-7) [85], which is a 7 item self-reported questionnaire that assesses symptoms of general anxiety according to a 4 point Likert-type scale, ranging from 0 (not at all) to 3 (nearly every day). These seven items assess (1) feeling nervous, anxious, or on edge; (2) being unable to stop or control worrying; (3) worrying too much about different things; (4) having trouble relaxing; (5) being restless; (6) becoming easily annoyed or irritable; and (7) feeling afraid as if something awful might happen. Items are summed to create a severity score ranging from 0 to 21. Scores above 10 are considered to be in the clinical range [85]. The scores of the GAD-7 are usually used as categorical variables due to their clinical relevance. In this study, we present only frequencies and ANOVA results based on the GAD-7 categories. Due to the relatively small sample size, we used the GAD-7 as a continuous variable to apply the SEM, because research has suggested that the GAD-7 can also be examined as a continuous variable [85]. To apply the SEM with categorical variables, a considerably larger sample in each category is needed. Unfortunately, the number of participants in each category was insufficient for the chosen statistical models.
Based on a validation study in the general population, the CFA results substantiated the one-dimensional structure of the GAD-7 and its factorial invariance for gender and age [88]. Research reported that the internal consistency α of the GAD-7 was 0.89, and intercorrelations with the PHQ-2 and the Rosenberg Self-Esteem Scale were r = 0.64 (p < 0.001) and r = −0.43 (p < 0.001) [88].

The MFI-20
To assess fatigue, we applied the Multidimensional Fatigue Inventory (MFI-20) [86]. The MFI-20 consists of 20 items categorized to five dimensions: general fatigue, physical fatigue, reduced activity, reduced motivation, and mental fatigue. For each scale, two items are oriented in the direction of fatigue, whereas the other two items are oriented in the opposite direction. Example items for the scale "general fatigue" are "I feel tired" and "I feel fit." The responses to each item are captured with a five-point Likert scale, ranging from 1 (yes, this is true) to 5 (no, this is not true). MFI-20 convergent validity was verified by correlating the hospital anxiety and depression scale (HADS) and the global quality of life scale of the EORTC QLQ-C30. Research has reported that the internal consistency α of the MFI-20 instrument was 0.72-0.87 [89].

The Learning Motivating Factors Questionnaire
To assess learning motivation, we applied the Learning Motivating Factors Questionnaire, developed by Law et al. [6,83]. This 19 item questionnaire measures factors that have a positive motivating effect on learning and covers several motivational variables listed below. The individual attitude and expectation subscale measures the student's attitude and expectation towards learning. The challenging goals subscale measures perceived challenging goals in learning. The clear direction subscale measures perceived specified direction in learning. The reward and recognition subscale measures perceived positive reinforcements, such as reward, appreciation, and encouragement. The punishment subscale measures the perceived negative reinforcement due to punishment. The social pressure and competition subscale measures perceived forces of pressure and competition from peers. The e-effect subscale measures the perceived effect of the e-learning setting. The response pattern followed a 6-point Likert scale, ranging from 1 (disagree very much) to 6 (agree very much). The validity of the construct was verified through oblique rotation exploratory factorial analysis. Research reported that the internal consistency α of the instrument was 0.95. The discriminant validity of each construct was checked using a multi-trait matrix [6].
In this study, for reliability analysis, Cronbach's alpha indexes were calculated. Cronbach alphas for the used instruments (the PHQ-9, the GAD-7, the MFI-20, the Learning Motivating Factors Questionnaire) in this research sample are presented in Table 1.

Statistical Analysis
For data analysis, we used SPSS v. 26.0 (IBM Corp., Armonk, NY, USA). The structural equation modeling (SEM), confirmatory factor analysis (CFA) models were conducted using AMOS v.26.0 (IBM Corp., Armonk, NY, USA) and JASP v. 0.14.1.0 (University of Amsterdam, Amsterdam, The Netherlands). Model fit was evaluated based on the CFI (Comparative Fit Index), the Normed Fit Index (NFI), the Tucker-Lewis coefficient (TLI), RMSEA (Root Mean Square Error of Approximation), and SRMR (Standardized Root Mean Square Residual), whereas the χ 2 was used for descriptive purposes only because it is highly sensitive to sample size [90]. The values higher than 0.90 for CFI, NFI, and TLI, and values lower than 0.08 for RMSEA and SRMR, were considered as indicative of a good fit [91]. In this research, we considered p-values less than 0.05 to be statistically significant [92].
The Therefore, we conducted a square root transformation (SQRT) of significantly negatively skewed variables to create normally distributed variables and conduct the CFA analyses.

Results
The frequencies of the self-reported depression categories in this sample are presented in Table 2. The frequencies of the self-reported anxiety categories in this sample are presented in Table 3. The frequencies of the self-reported general fatigue categories in this sample are presented in Table 4. The means, standard deviations, and correlations between the MFI-20 subscales in this study are reported in Table 5. The means, standard deviations, and correlations between the Learning Motivating Factors Questionnaire subscales in this study are reported in Table 6. The means, standard deviations, and correlations between the PHQ-9 and the GAD-7 scales in this study are reported in Table 7. To test hypothesis 1 (H1), assuming that lower scores of depression relate to higher scores of learning motivating factors, we compared the scores of learning motivating factors based on the PHQ-9 categories (Table 8). A one-way ANOVA was conducted to compare the effect of depression (PHQ-9) on learning motivating factors. We analyzed learning motivating factors in five groups: minimal depression, mild depression, moderate depression, moderately severe depression, and severe depression. An analysis of variance revealed that the effect of depression on individual attitude and expectation was significant, F (4, 387) = 3.324, p = 0.011. A Tukey post hoc test revealed that the scores of individual attitude and expectation were significantly higher in the minimal depression group (4.859, ±0.783) in comparison to the moderately severe depression group (4.403, ±0.904) (p = 0.018).
An analysis of variance revealed that the effect of depression on clear direction was also significant, F (4, 387) = 6.884, p < 0.001. A Tukey post hoc test revealed that the scores of clear direction were significantly higher in the minimal depression group (5.172, ±0.737) in comparison to the moderately severe depression (4.612, ±0.681) (p < 0.001) or severe depression (4.583, ±1.057) (p = 0.001) groups. Additionally, the scores of clear direction were significantly higher in the mild depression group (5.015, ±0.719) in comparison to the moderately severe depression group (4.612, ±0.681) (p = 0.014) and the severe depression group (4.583, ±1.057) (p = 0.033).
An analysis of variance revealed that the effect of depression on social pressure and competition was significant, F (4, 387) = 2.657, p = 0.033. A Tukey post hoc test revealed that the scores of social pressure and competition were significantly higher in the mild depression group (3.604, ±1.176) in comparison to the moderately severe depression group (2.975, ±1.198) (p = 0.016).
However, there were no significant differences between the scores of reward and recognition in the depression groups F (4, 387) = 0.972, p = 0.423. In addition, there were no significant differences between the scores of punishment in the depression groups, F (4, 387) = 1.901, p = 0.110.
Furthermore, to test hypothesis 2 (H2), which presumed that lower anxiety scores relate to higher scores of intrinsic, but not extrinsic, learning motivating factors, we compared the scores of learning motivating factors based on the GAD-7 categories (Table 9). A one-way ANOVA was conducted to compare the effect of anxiety (GAD-7) on learning motivating factors. We analyzed learning motivating factors in four groups: minimal anxiety, mild anxiety, moderate anxiety, and severe anxiety. An analysis of variance revealed that the effect of severe anxiety on challenging goals was significant, F (3, 401) = 8.345, p < 0.001. A Tukey post hoc test revealed that the scores of challenging goals were significantly higher in the minimal anxiety group (4.604, ±1.020) in comparison to the moderate anxiety group (4.155, ±0.946) (p = 0.018) and the severe anxiety group (3.782, ±1.154) (p < 0.001). Additionally, challenging goals were significantly higher in the mild anxiety group (4.329, ±1.112) in comparison to the severe anxiety group (3.782, ±1.154) (p = 0.011).
An analysis of variance revealed that the effect of severe anxiety on clear direction was significant, F (3, 401) = 3.449, p = 0.017. Similarly, the effect of severe anxiety on social pressure and competition was significant, F (3, 401) = 2.784, p = 0.041. A Tukey post hoc test revealed that the scores of social pressure and competition were significantly higher in the minimal anxiety group (3.525, ±1.127) in comparison to the severe anxiety group (3.010, ±1.187) (p = 0.047). Additionally, scores of social pressure and competition were significantly higher in the mild anxiety group (3.564 ± 1.327) in comparison to the severe anxiety group (3.010, ±1.187) (p = 0.030).
Next, to test hypothesis 3 (H3), assuming that lower scores of general fatigue relate to higher scores of learning motivating factors, we compared the scores of learning motivating factors based on the general fatigue (MFI-20) categories (Table 10). A one-way ANOVA was conducted to compare the effect of fatigue on learning motivating factors. We analyzed learning motivating factors in four groups: minimal fatigue, mild fatigue, moderate fatigue, and severe fatigue. An analysis of variance revealed that the effect of general fatigue on individual attitude and expectation was significant, F (3, 402) = 2.628, p = 0.050. A Tukey post hoc test revealed that the scores of individual attitude and expectation were significantly higher in the moderate fatigue group (4.801, ±0.831) in comparison to the severe fatigue group (4.481, ±1.023) (p = 0.034).
An analysis of variance revealed that the effect of general fatigue on challenging goals was significant, F (3, 402) = 7.476, p < 0.001. A Tukey post hoc test revealed that the scores of challenging goals were significantly higher in the minimal fatigue group (4.778, ±1.112) in comparison to the moderate fatigue group (4.306, ±0.918) (p = 0.036) and the severe fatigue group (3.921, ±1.096) (p < 0.001). Additionally, the scores of challenging goals were significantly higher in the mild fatigue group (4.459, ±1.058) in comparison to the severe fatigue group (3.921, ±1.096) (p = 0.003).
An analysis of variance also revealed that the effect of general fatigue on social pressure and competition was significant, F (3, 402) = 6.162, p < 0.001. A Tukey post hoc test revealed that the scores of social pressure and competition were significantly higher in the mild fatigue group (3.687, ±1.255) in comparison to the severe fatigue group (3.053, ±1.270) (p = 0.001). Additionally, the scores of social pressure and competition were significantly higher in the moderate fatigue group (3.585, ±1.057) in comparison to the severe fatigue group (3.053, ±1.270) (p = 0.006).
Next, to test hypothesis 4 (H4), presuming that extrinsic learning motivating factors predict increased anxiety, and intrinsic learning motivating factors predict decreased anxiety, we conducted a multiple linear regression using anxiety (GAD-7) as the criterion and learning motivating factors as predictors (Table 11) in groups of respondents, participating and not participating in e-learning-based computer programming courses. A multiple linear regression model was calculated to predict anxiety based on learning motivating factors in groups of respondents participating and not participating in e-learning-based computer programming courses. A significant regression equation was found only in group of respondents participating in computer programming courses, F (6, 179) = 5.508, p < 0.001), with an R 2 = 0.156. Respondents' predicted anxiety was equal to 15.287 − 1.692 (challenging goals) points. Anxiety decreased −1.692 points for each challenging goals motivation point. Challenging goals (B = −1.692, p < 0.000) contributed significantly to the model and were significant predictors of diminished anxiety.
Furthermore, to test hypothesis 5 (H5), which assumed that extrinsic learning motivating factors predict increased depression, and intrinsic learning motivating factors predict decreased depression, we conducted a multiple linear regression using depression (PHQ-9) as the criterion and learning motivating factors as predictors (Table 12).  Additionally, to test hypothesis (H6), assuming that extrinsic learning motivating factors, anxiety, and depression predict increased general and mental fatigue, and intrinsic learning motivating factors predict decreased general and mental fatigue, we primarily conducted a multiple linear regression using general fatigue (MFI-20) as the criterion and learning motivating factors, anxiety, and depression as predictors (Table 13).  Then, we conducted a multiple linear regression using mental fatigue as the criterion and learning motivating factors, anxiety, and depression as predictors (Table 14).  Based on the literature review and previous analyses that indicate the differences and impact of several learning motivation factors, to test hypothesis 7 (H7), assuming that associations exist between learning motivating factors, anxiety, depression, and fatigue, but they differ between participants and non-participants of e-learning-based computer programming courses, we first created a model of associations between the learning motivating factors (challenging goals, punishment, social pressure, and competition), anxiety, depression, and fatigue in the group of participants of e-learning-based computer programming courses. Standardized results of the model are presented in Figure 1   Scalar estimates of the model of associations between learning motivating factors, anxiety, depression, and fatigue in the sample of participants of e-learning-based computer programming courses are displayed in Table 15. Furthermore, we created a model of associations between some learning motivating factors (challenging goals, punishment, social pressure, and competition), anxiety, depression, and fatigue in the group of those who did not participate in e-learning-based computer programming courses. Standardized results of the model are presented in Figure 2. To assess the model fit, the Comparative Fit Index (CFI), the Tucker-Lewis coefficient (TLI), and Root Mean Square Error of Approximation (RMSEA) were used. Findings revealed that the fit of the model was acceptable, χ 2 = 563.004; df = 288; p < 0.001; CFI = 0.908; TLI = 0.910; RMSEA = 0.061 [0.054-0.069].
Scalar estimates of the model of associations between learning motivating factors, anxiety, depression, and fatigue in the sample of non-participants of e-learning-based computer programming courses are displayed in Table 16.
Furthermore, based on the results of multiple regression analysis, structural equation modeling (SEM) was conducted to analyze the associations between challenging goals, mental fatigue, depression, and anxiety in the group of participants of e-learning-based computer programming courses. The path diagram is displayed in Figure 3.  Scalar estimates of the model of associations between learning motivating factors, anxiety, depression, and fatigue in the sample of non-participants of e-learning-based computer programming courses are displayed in Table 16. Table 16. Scalar estimates of the model of associations between learning motivating factors, anxiety, depression, and fatigue in the sample of respondents who did not participate in e-learning-based computer programming courses.  Furthermore, based on the results of multiple regression analysis, structural equation modeling (SEM) was conducted to analyze the associations between challenging goals, mental fatigue, depression, and anxiety in the group of participants of e-learning-based computer programming courses. The path diagram is displayed in Figure 3.  To examine alternative models of associations between challenging goals, mental fatigue, depression, anxiety, and assess differences between participants and non-participants of e-learning-based computer programming courses, we conducted structural equation modelling with two sample groups. Respondents of group 1 were not participating in e-learning-based program, and respondents of group 2 were participating in computer programming courses. The diagram of the path is displayed in Figure 4. The results of SEM analysis are presented in Table 18. SEM analysis showed that the motivational factor challenging goals statistically significantly predicts mental fatigue, and mental fatigue statistically significantly predicts depression and anxiety in both sample groups. In Group 2, challenging goals explains 25.7% of the variance of mental fatigue, and mental fatigue explains 44.1% of the variance of depression and 31.1% of the variance of anxiety. In group 1, challenging goals explains 5.5% of the variance of mental fatigue, and mental fatigue explains 7.7% of the variance of depression and 9.6% of the variance of anxiety. Findings revealed that the fit of the model was acceptable, To examine alternative models of associations between challenging goals, mental fatigue, depression, anxiety, and assess differences between participants and nonparticipants of e-learning-based computer programming courses, we conducted structural equation modelling with two sample groups. Respondents of group 1 were not participating in e-learning-based program, and respondents of group 2 were participating in computer programming courses. The diagram of the path is displayed in Figure 4. The results of SEM analysis are presented in Table 18. SEM analysis showed that the motivational factor challenging goals statistically significantly predicts mental fatigue, and mental fatigue statistically significantly predicts depression and anxiety in both sample groups. In Group 2, challenging goals explains 25.7% of the variance of mental fatigue, and mental fatigue explains 44.1% of the variance of depression and 31.1% of the variance of anxiety. In group 1, challenging goals explains 5.5% of the variance of mental fatigue, and mental fatigue explains 7.7% of the variance of depression and 9.6% of the variance of anxiety. Findings revealed that the fit of the model was acceptable,

Discussion
This study was the first to explore associations between depression, anxiety, fatigue, and learning motivating factors in e-learning-based education, including computer programming learning, during the COVID-19 pandemic. The examination of learning motivation was based on a model of extrinsic and intrinsic learning motivating factors, developed by Law et al. [6]. This research was conducted during the COVID-19 pandemic. The context of the COVID-19 quarantine can be characterized by the rapid implementation of e-learning [93], which helped us compare the motivating factors in e-learning-based computer programming education and e-learning in education other than computer programming. The pandemic context can also be characterized by the statistically significant increase in the rates of anxiety and depression [65,70,[72][73][74], chronic fatigue [66], and psychological distress [67]. Due to these conditions, the research questions were extremely important, because previous studies indicated that the states of severe anxiety and depression relate to dramatically diminished e-learning motivation, or even its absence [76].

Lower Scores of Depression Partially Relate to Higher Scores of Learning Motivating Factors
In this study, we assumed that lower scores of depression relate to higher scores of learning motivating factors. Thus, we compared the scores of learning motivating factors based on the PHQ-9 categories. The results partially confirmed this hypothesis and showed that the scores of the motivating factor individual attitude and expectation were significantly higher in the minimal depression group in comparison to the moderately severe depression group. The scores of challenging goals were significantly higher in the minimal depression group in comparison to the mild depression, moderate depression, moderately severe depression, and severe depression groups, and the scores of challenging goals were significantly higher in the mild depression group in comparison to the moderately severe depression group. The scores of clear direction were significantly higher in the minimal depression group in comparison to the moderately severe depression or severe depression groups, and the scores of clear direction were significantly higher in the mild depression group in comparison to the moderately severe depression group and the severe depression group. The scores of social pressure and competition were significantly higher in the mild depression group in comparison to the moderately severe depression group. However, there were no significant differences in the scores of reward and recognition and the scores of punishment in the depression groups. To summarize, the scores of individual attitude and expectation, challenging goals, clear direction, social pressure, and competition varied between the depression categories. However, no significant differences were identified in the scores of reward and recognition and punishment in the different depression-level groups. These results align with previous research suggesting that depression is associated with learning difficulties, both behavioral and neural, partly due to impaired generation and updating of outcome predictions [94]. Moreover, these findings support studies implying that learning motivation can be impacted by depression [95][96][97][98].

Lower Anxiety Scores Partially Relate to Higher Scores of Intrinsic Learning Motivating Factors
Furthermore, we presumed that lower anxiety scores relate to higher intrinsic, but not extrinsic, motivating factors. Hence, we compared the scores of learning motivating factors based on the GAD-7 categories. The results partially confirmed this hypothesis and showed that the scores of challenging goals were significantly higher in the minimal anxiety group than the moderate anxiety and severe anxiety groups. Challenging goals were significantly higher in the mild anxiety group in comparison to the severe anxiety group; the scores of social pressure and competition were significantly higher in the minimal anxiety group in comparison to the severe anxiety group, and the scores of social pressure and competition were significantly higher in the mild anxiety group in comparison to the severe anxiety group. However, there were no significant differences in individual attitude and expectation, reward and recognition, and punishment scores between different anxiety groups. To summarize, in this study, the scores of challenging goals, and social pressure and competition, significantly varied between different anxiety categories. However, no significant differences were identified in the scores of individual attitude and expectation, reward and recognition, and punishment between different anxiety-level groups. These findings partially contradict some previous results evidencing that anxiety is unrelated to learning outcomes and motivation in computer programming learning [37][38][39]. However, the findings support those of other studies, which demonstrated the impact of anxiety on learning motivation [76,99,100], and the impact of learning motivation on emotions [25][26][27].

Lower Scores of General Fatigue Partially Relate to Higher Scores of Learning Motivating Factors
Next, we assumed that lower scores of general fatigue relate to higher scores of learning motivating factors. Therefore, we compared the scores of learning motivating factors based on the general fatigue (MFI-20) categories. The results partially confirmed this hypothesis and showed that the scores of individual attitude and expectation were significantly higher in the moderate fatigue group in comparison to the severe fatigue group. The scores of challenging goals were significantly higher in the minimal fatigue group in comparison to the moderate fatigue group and the severe fatigue group, and the scores of challenging goals were significantly higher in the mild fatigue group in comparison to the severe fatigue group. The scores of social pressure and competition were significantly higher in the mild fatigue group in comparison to the severe fatigue group, and the scores of social pressure and competition were significantly higher in the moderate fatigue group in comparison to the severe fatigue group. However, there were no significant differences in the scores of clear direction, reward and recognition, and punishment between different self-reported fatigue-level groups. To summarize, this study demonstrated that the scores of individual attitude and expectation, challenging goals, social pressure, and competition varied across different fatigue categories, but no significant differences were identified in the scores of clear direction, reward and recognition, and punishment between different fatigue level-groups. These results partially confirm the findings of other authors, evidencing associations between fatigue and learning motivation [101][102][103][104][105][106]. However, it is unclear whether learners' fatigue was related to learning burnout or the psychologically challenging COVID-19 pandemic, and why some motivational factors, for example, striving for reward and recognition or fear of punishment, did not contribute to the increased fatigue.

Learning Motivating Factors Partially Predict Anxiety Level
Next, we presumed that extrinsic learning motivating factors predict increased anxiety, and intrinsic learning motivating factors predict decreased anxiety. Thus, we conducted a multiple linear regression using anxiety (GAD-7) as the criterion and learning motivating factors as predictors in groups of respondents participating and not participating in e-learning-based computer programming courses. The results partially confirmed this hypothesis and showed that in the group of respondents participating in computer programming courses, challenging goals was a significant predictor of diminished anxiety. To summarize, this study showed that intrinsic the motivating factor challenging goals predicts decreased anxiety in a group of participants of e-learning-based computer programming courses. These findings are, to some extent, in line with those of previous research, suggesting that intrinsic learning motivation is related to better learning outcomes and higher satisfaction with the learning process, whereas extrinsic learning motivation is related to higher stress levels [35,36,67]. However, it is partially clear why only one intrinsic motivational factor (challenging goals) contributed to the decreased anxiety [107], and why this pattern was not apparent in the group of other e-learners, which would be in line with many other studies [33,[107][108][109][110][111].

Learning Motivating Factors Partially Predict Depression Level
Furthermore, we assumed that extrinsic learning motivating factors predict increased depression, and intrinsic learning motivating factors predict decreased depression. Thus, we conducted a multiple linear regression using depression (PHQ-9) as the criterion and learning motivating factors as predictors in groups of respondents participating and not participating in e-learning-based computer programming courses. To a small extent, The results confirmed this hypothesis and showed that, in the group of respondents who did not participate in computer programming courses, the challenging goals factor was a significant predictor of diminished depression. However, the motivating intrinsic factor of challenging goals and the extrinsic factor of clear direction were significant predictors of diminished depression in the group of participants of e-learning-based computer programming courses. To summarize, this research demonstrated that the extrinsic motivating factor of clear direction and the intrinsic motivating factor of challenging goals predict decreased depression in the group of participants of e-learning-based computer programming courses, whereas the intrinsic motivating factor of challenging goals predicts decreased depression in the group of other e-learners. These findings are moderately supported by previous research suggesting the benefits of some intrinsic and extrinsic motivating factors [32,33,107,[111][112][113]. However, the added value of clear direction and challenging goals for computer programming learners needs further investigation.

Learning Motivating Factors, Anxiety, and Depression Partially Predict Fatigue
Additionally, we assumed that extrinsic learning motivating factors, anxiety, and depression predict increased general and mental fatigue, and intrinsic learning motivating factors predict decreased general and mental fatigue. Thus, we conducted a multiple linear regression using general and mental fatigue (MFI-20) as the criteria and learning motivating factors, anxiety, and depression as predictors. The results, to some extent, confirmed this hypothesis and showed that individual attitudes and expectations, anxiety, and depression predicted higher general fatigue of respondents who did not participate in e-learning-based computer programming courses. Interestingly, the computer programming learners' intrinsic motivating factor of individual attitude and expectation predicted diminished general fatigue; however, clear direction, reward and recognition, anxiety, and depression were sig-nificant predictors of increased general fatigue of participants of e-learning-based computer programming courses. Furthermore, anxiety and depression were significant predictors of mental fatigue of respondents not participating in computer programming courses. Remarkably, the computer programming learners' intrinsic motivating factor of challenging goals and their extrinsic motivating factor of punishment predicted diminished mental fatigue; however, social pressure and competition, and depression, were significant predictors of increased mental fatigue of participants of e-learning-based computer programming courses. To summarize, this study found that the intrinsic motivating factor of individual attitude and expectation predicted diminished general fatigue, and clear direction, reward and recognition, anxiety, and depression predicted increased general fatigue of participants of e-learning-based computer programming courses. Individual attitudes and expectations, anxiety, and depression predicted higher general fatigue of non-participants; the computer programming learners' intrinsic motivating factor of challenging goals and their extrinsic motivating factor of punishment predicted diminished mental fatigue, whereas the mental fatigue of other e-learners was predicted only by anxiety and depression. Therefore, this research modestly contributed to a better understanding of the complexity of associations between intrinsic and extrinsic learning motivating factors, anxiety, depression, and fatigue, as analyzed by other researchers [114][115][116][117][118][119]. However, it is unclear why computer programming e-learners' individual attitude and expectation predicted diminished general fatigue, whereas other e-learners' individual attitude and expectation predicted an increase in general fatigue. This question needs further investigation.

Associations between the Study Variables Partially Differ in the Compared Groups
Based on a literature review and previous analyses, which suggested the differences and impact of several learning motivation factors, we assumed that associations exist between learning motivating factors, anxiety, depression, and fatigue, and that these associations differ between participants and non-participants of e-learning-based computer programming courses. Thus, we tested several models of associations between the study variables. This study partially confirmed the hypothesis and identified several possible paths and models of associations between learning motivating factors (challenging goals, punishment, social pressure, and competition), anxiety, depression, and fatigue in groups of participants and non-participants of e-learning-based computer programming courses. However, although some differences in associations between the groups were identified, the SEM analyses demonstrated that the intrinsic motivating factor of challenging goals statistically significantly predicts lower mental fatigue, and mental fatigue statistically significantly predicts depression and anxiety in both groups of participants and nonparticipants of e-learning-based computer programming courses. To summarize, this study demonstrated that associations between learning motivating factors, anxiety, depression, and fatigue differ between participants and non-participants of e-learning-based computer programming courses. However, the intrinsic motivating factor of challenging goals statistically significantly predicts lower mental fatigue, whereas mental fatigue statistically significantly predicts depression and anxiety in both sample groups. These findings are consistent with many previous studies that suggest the added value of intrinsic learning motivation [33,99,[107][108][109][110][111]. In the future, it will be essential that the identification of the associations between learning motivating factors and the positive states receives increased attention from researchers [120,121].

Limitations and Future Directions
Several limitations to this study can be mentioned. First, because the mental health issues in e-learning-based education are under-researched, this study lacks a solid theoretical basis. Second, because the objective health status was not measured, bias may have occurred due to the use of self-reported measures only, and the omission of objective indicators. Third, the findings should be regarded with caution, considering that the data were collected online. Fourth, the research samples were not representative but random, suggesting the necessity to analyze representative samples of e-learners; thus, generalizations should be made with caution. Fifth, although the sample size satisfied the minimal requirements for the applied statistical models, and the data fit was acceptable, the results should be regarded cautiously due to the relatively small sample size. Sixth, some possible theoretical bias may have occurred in the selected model. Furthermore, this study was conducted in Lithuania, and the results may reflect the specifics of this area, suggesting the necessity to analyze the impact of cultural factors, considering the more specific aspects of each culture. Finally, the findings suggest a necessity for longitudinal or experimental research design, because, based on the data obtained, it is possible to only identify the existence of significant relationships among the examined variables. Thus, the conclusions should be regarded with caution, especially regarding causality. Although this study found that higher scores of individual attitude and expectation, challenging goals, clear direction, social pressure, and competition are associated with better mental health, the causality issues need further investigation, because reverse causality is also likely to occur.

Theoretical Implications
From a theoretical perspective, this study was the first of its kind to explore the associations between depression, anxiety, fatigue, and learning motivating factors in elearning-based education, including computer programming learning, during the COVID-19 pandemic. Many learning motivation models exist, but this study was based on a theory of extrinsic and intrinsic learning motivating factors, developed by Law et al. [6]. Although the findings revealed the complexity of the relations between learning motivating factors, anxiety, depression, and fatigue, numerous questions remained unanswered. It is unclear whether learners' fatigue relates to learning burnout, or was a consequence of the psychologically challenging COVID-19 pandemic. In addition, it is unclear why striving for reward and recognition or fear of punishment did not increase fatigue. Furthermore, it is only partially clear why only one intrinsic motivational factor (challenging goals) contributed to the decreased anxiety and why this pattern was not apparent in the group other than computer programming e-learners. The emotional benefits of clear direction and challenging goals for computer programming learners also need further investigation. It is also unclear why computer programming e-learners' individual attitude and expectation predicted diminished general fatigue, whereas other e-learners' individual attitude and expectation predicted an increase in general fatigue. In the future, it would be interesting to identify the associations between intrinsic and extrinsic learning motivating factors and the positive states such as flow or zest, which have recently received the increased attention of researchers [120,121].

Practical Implications
The COVID-19 quarantine forced the rapid implementation of e-learning [93]. The psychologically challenging pandemic also increased the rates of anxiety, depression [65,70,[72][73][74], and fatigue [66]. Because severe anxiety and depression states relate to dramatically diminished e-learning motivation, or even its absence [76], it was important to identify e-learning motivating factors related to mental health under the challenging psychological circumstances of the COVID-19 pandemic. Because computer programming skills are a core competency that professionals are expected to possess in the era of rapid technology development [1], it was also essential to identify the specific factors related to computer programming learning. This research revealed that mental health, namely, depression, anxiety, and fatigue, are related to the scores of intrinsic and extrinsic learning motivating factors. In addition, the motivating factor of challenging goals was strongly related to better mental health in groups of participants and non-participants of e-learningbased computer programming courses. This implies that e-learners' good mental health is a target for both health and education policy makers, health professionals, and educators.

Conclusions
This study aimed to identify associations between depression, anxiety, fatigue, and learning motivating factors in participants and non-participants of e-learning-based computer programming courses. It was found that the scores of individual attitude and expectation, challenging goals, clear direction, social pressure, and competition varied between the depression categories. However, no significant differences were identified in the scores of reward and recognition, and punishment, in the different depression-level groups. The scores of challenging goals, and social pressure and competition, significantly varied between the anxiety categories, but no significant differences were identified in the scores of individual attitude and expectation, reward and recognition, and punishment between different anxiety-level groups. The scores of individual attitude and expectation, challenging goals, and social pressure and competition varied across different fatigue categories, but no significant differences were identified in the scores of clear direction, reward and recognition, and punishment between different fatigue-level groups. The intrinsic motivating factor of challenging goals predicted decreased anxiety in a group of participants of e-learning-based computer programming courses. The extrinsic motivating factor of clear direction and the intrinsic motivating factor of challenging goals predicted decreased depression in a group of participants of e-learning-based computer programming courses, and the intrinsic motivating factor of challenging goals predicted decreased depression in a group of other e-learners. This study found that the intrinsic motivating factor of individual attitude and expectation predicted diminished general fatigue, and clear direction, reward and recognition, anxiety, and depression predicted increased general fatigue of participants of e-learning-based computer programming courses. Individual attitude and expectation, anxiety, and depression predicted higher general fatigue of non-participants. The computer programming learners' intrinsic motivating factor of challenging goals and their extrinsic motivating factor of punishment predicted diminished mental fatigue, whereas the mental fatigue of other e-learners was predicted only by anxiety and depression. Associations between learning motivating factors, anxiety, depression, and fatigue differed between participants and non-participants of e-learning-based computer programming courses. However, the intrinsic motivating factor of challenging goals statistically significantly predicted lower mental fatigue, whereas mental fatigue statistically significantly predicted depression and anxiety in both sample groups. Nevertheless, the results of this particular study should be regarded with caution due to the relatively small sample size, the possible biases in the selected theoretical and statistical models, the research design, and other limitations.